LoRaWAN vs NB-IoT vs LTE-M: Choosing the Right Network for Asset Tracking

Every asset-tracking project starts with the same question and the same wrong answer. The question: which low-power wide-area network should we use? The wrong answer: pick one and standardize. The right answer is that LoRaWAN, NB-IoT, and LTE-M each win specific scenarios, and a serious fleet usually ends up using two of them. This article compares the three on the dimensions that actually drive design – range, throughput, power, coverage, geolocation, and total cost of ownership over five years for a thousand devices. By the end you should know which network fits your asset, your geography, and your business model.

The Three Networks at a Glance

All three are LPWAN technologies designed for devices that send small payloads infrequently and need to run for years on a battery. They differ in spectrum, modulation, ownership model, and ecosystem.

• LoRaWAN uses unlicensed sub-GHz ISM bands (EU868, US915, AS923) with chirp spread spectrum modulation. Networks can be public (operators like Helium, The Things Network, national operators) or private (you run your own gateways). The MAC and application layers are open standards.
• NB-IoT (LTE Cat NB1/NB2) is a 3GPP standard riding on licensed cellular spectrum, optimized for deep-indoor coverage and very low data rates. Always operator-provided. Uses about 200 kHz of bandwidth.
• LTE-M (LTE Cat M1) is the other 3GPP LPWAN standard, with higher throughput than NB-IoT, support for mobility (handover), and limited voice support. Also always operator-provided.

The big architectural fork is licensed versus unlicensed. NB-IoT and LTE-M depend on a mobile network operator owning the spectrum and the infrastructure. LoRaWAN can be deployed entirely under your control if you accept the gateway and backhaul cost. That single difference cascades into every other tradeoff that follows.

Side-by-Side Comparison

The numbers above are reference points from our own deployments and from current operator pricing across the EU and North America in 2026. Pricing varies wildly by volume; the floor for cellular IoT plans drops fast at fleet sizes above 10,000 devices, and large enterprise contracts can move the per-device-per-month cost by an order of magnitude.

Use Cases That Each Network Wins

LoRaWAN Wins When You Own the Site

Hotels, marinas, factories, ports, vineyards, ski resorts. Any environment where you control real estate, can mount a few gateways, and have hundreds to thousands of low-bandwidth sensors. The economics are unbeatable: after the gateway capex, the marginal cost per device per year is zero. Battery lifetimes routinely exceed five years on a single AA-class cell.

For a luxury resort, we have deployed 1,200 LoRaWAN sensors – door contacts, mini-bar weights, environmental probes, asset tags on guest equipment – on three indoor gateways and one outdoor gateway covering the grounds. Connectivity opex: zero. Replacement batteries every five to seven years. The same fleet on cellular would have cost approximately $14,000 per year in connectivity alone.

NB-IoT Wins for Deep-Indoor, Stationary, Low-Bandwidth Cases

Smart meters in basements. Parking sensors under asphalt. Water quality sondes in wells. Anything that needs to send a few hundred bytes a day from somewhere a normal cell signal cannot reach. NB-IoT’s deep-indoor link budget (around 164 dB MCL) penetrates concrete in ways LTE-M and LoRaWAN cannot.

The catch is mobility. NB-IoT was not designed for moving devices and handles cell reselection slowly. If your asset moves while transmitting, expect missed messages. For stationary assets in challenging RF environments, NB-IoT is unmatched.

LTE-M Wins for Mobility and Wider Coverage Without Site Build

Vehicle tracking, container tracking across borders, mobile medical devices, pet trackers, anything that crosses cells while in motion. LTE-M’s full handover support and lower latency make it the only LPWAN choice for moving assets that need to maintain a session. It is also the only 3GPP LPWAN with credible roaming agreements across most of the world.

For a marine bunker fuel logistics customer, we used LTE-M on harbor tugs and barges that transit between EU ports. Roaming is handled by a single global SIM provider, geolocation is GPS plus cell fallback, and the whole fleet is visible in our standard fleet management dashboards.

Hybrid Strategies Are Often Right

The interesting designs are hybrid. A few patterns we use:

• LoRaWAN on-site, LTE-M in transit: assets carry both radios. On a customer’s property, they speak LoRaWAN and pay nothing. Off-property, they fall back to LTE-M. The radio choice is a runtime decision based on RSSI to the known LoRaWAN gateways.
• NB-IoT primary, LTE-M failover: smart-city devices that mostly sit in deep-indoor NB-IoT coverage but occasionally need higher throughput for firmware updates.
• Satellite topup: a few of our marine customers add a low-bandwidth satellite modem (Iridium SBD or the new direct-to-cell offerings) for vessels that go out of LTE-M range, reserved for safety-critical telemetry.

Multi-radio devices cost more in BOM and firmware complexity, but for high-value assets that move between connectivity zones, the math usually favors hybrid. The decision logic belongs in firmware that you control end-to-end, which is one reason we recommend keeping radio selection out of vendor SDKs and into your own state machine.

Modems and Antennas

Module choice is more consequential than people think. A short reference of what we currently spec:

• LoRaWAN end nodes: Semtech SX1262 or LR1110 (the LR1110 integrates GNSS and Wi-Fi sniffing for geolocation). Murata CMWX1ZZABZ for an SoC-style module.
• NB-IoT/LTE-M: Quectel BG95 family for cost, u-blox SARA-R5 for a more deterministic AT firmware and longer support lifecycle, Sequans Monarch for the lowest power.
• Combo modules: u-blox SARA-R510M8S adds GNSS to LTE-M; Quectel BG95-M3 supports NB-IoT, LTE-M, and GNSS in one part.

Antennas are where most projects fail. A bad PCB antenna costs you 6-10 dB versus a properly tuned external one – that is the difference between five-year and two-year battery life, or the difference between coverage and a return visit. We spec proper matching networks on every PCB design, validate with a VNA, and budget for a real RF anechoic test if the device ships in volume. If the antenna is an afterthought, the radio link budget will be too.

Geolocation: GPS, Cell, LoRa TDoA, Wi-Fi Sniffing

Asset tracking implies position. The right answer depends on accuracy needs and power budget.

• GNSS (GPS/Galileo/GLONASS/BeiDou): 3-10 m accuracy outdoors, useless indoors, drains 20-50 mA while acquiring fix. Use assisted-GPS (A-GNSS) over the cellular link to cut acquisition time from minutes to seconds and save battery.
• Cell-tower triangulation: 100-2000 m accuracy depending on tower density. Free (computed from the same cell info you already have) but coarse. Useful as a fallback when GNSS cannot fix.
• LoRa TDoA: time-difference-of-arrival across multiple gateways gives 50-200 m accuracy without any GPS hardware. Requires three or more gateways with synchronized timestamps and works only in your private network.
• Wi-Fi BSSID sniffing: scan nearby BSSIDs and look them up against a geolocation database (Google, Combain). 20-100 m accuracy in dense urban areas. Modern modules like the LR1110 can do this without a dedicated Wi-Fi chip.

For most outdoor mobile assets, GNSS plus A-GNSS over LTE-M is the right baseline, with cell fallback for bad sky conditions. For containers in warehouses and ports, Wi-Fi sniffing plus cell typically beats GNSS. For sensors on a private LoRaWAN site, TDoA is essentially free coarse positioning.

Private LoRaWAN vs Public Operators

The case for private LoRaWAN is control. You own the gateways, you control the join server, you set the data rate and channel plan, and you do not pay per message. The case against is that you are responsible for everything – gateway uptime, backhaul, security, regulatory compliance.

For sites with more than 50 devices that you control physically, private wins almost always. For genuinely distributed deployments with handfuls of devices per site, public networks (national operators or Helium-style community networks) make sense despite the per-message cost. We provision both kinds through the same backend pipeline so the application layer never has to know which network a packet came from. Standard integration patterns hide the gateway boundary.

Five-Year TCO for 1000 Devices

Numbers below assume 1000 devices, 24 messages per day each, 50-byte payloads, 5-year horizon, EU pricing in 2026. Hardware BOM at the module level only – your finished-good cost includes much more.

• Private LoRaWAN: Modules $4 x 1000 = $4,000. Three gateways at $1,200 each = $3,600. Backhaul and hosting $200/month x 60 = $12,000. Battery replacements assume one cycle = $5,000. Total: ~$24,600.
• NB-IoT public: Modules $7 x 1000 = $7,000. Connectivity $0.50/device/month x 1000 x 60 = $30,000. Battery replacements two cycles = $10,000. Total: ~$47,000.
• LTE-M public: Modules $10 x 1000 = $10,000. Connectivity $1.20/device/month x 1000 x 60 = $72,000. Battery replacements three cycles = $15,000. Total: ~$97,000.

The TCO ratio is roughly 1 : 2 : 4. That spread justifies real engineering effort to push as much of your fleet as possible onto private LoRaWAN where the use case allows. It also justifies hybrid radios on the high-value subset that genuinely needs LTE-M mobility. Fleet managers who treat radio choice as a per-asset decision rather than a global default routinely cut connectivity opex in half.

Regulatory Notes

Unlicensed bands have duty-cycle limits that bite. EU868 imposes a 1% duty cycle on the most common sub-bands, which translates to roughly 36 seconds of airtime per hour. With SF12 long-range frames eating most of that budget, the practical message rate per device is very limited. US915 has no duty cycle but enforces frequency hopping. Always model duty cycle at design time; it kills more LoRaWAN deployments than RF coverage does.

Cellular IoT inherits the regulatory regime of the host network, so you mostly do not worry about it – but you do need to confirm that the operator supports your chosen radio in every country you ship to. NB-IoT availability is patchy in North America; LTE-M is now ubiquitous in EU and NA but spotty elsewhere. The 2G and 3G sunset has finally cleared most spectrum confusion, but always check operator commitments to LPWAN bands before locking in a module.

Putting It Together

Pick LoRaWAN when you own the site. Pick NB-IoT when the asset is stationary, deep indoors, and ships only tiny payloads. Pick LTE-M when the asset moves and crosses cells. Combine them when the asset does multiple of those. And design the radio decision into the firmware state machine, not into the procurement spec, so you can shift fleets between networks as economics evolve. The same goes for the cloud side – the data pipeline downstream of the radio should look the same whether the bytes arrived via gateway, eNodeB, or satellite, which is exactly the abstraction our industrial IoT stack is built around.

If you are mapping out a connectivity strategy for a new asset-tracking product, we design these end-to-end – from antenna and PCB through firmware, modem certification, and cloud. Start with our connected devices service and we will work backwards from your assets to the right radio mix.